Biomedical Engineering Reference
In-Depth Information
Ham and Han (1996) investigated the use of fuzzy ART mapping
(ARTMAP) to classify cardiac arrhythmias based on the QRS complex. The
system was tested on the data from the MIT-BIH database where cardiac
beats for normal and abnormal QRS complexes were extracted, scaled, and
Hamming windowed. After bandpass filtering, a sequence of 100 samples for
each QRS segment was obtained and two conditions analyzed; normal and
abnormal PVC. From each of these sequences, two linear predictive coding
(LPC) coecients were generated using Burg's maximum entropy method.
These two LPC coecients were utilized with the mean-square value of the
QRS complex segment as features for each condition to train and test a
fuzzy ARTMAP NN. The test results indicated that the fuzzy ARTMAP NN
could classify cardiac arrhythmias with greater than 99% specificity and 97%
sensitivity.
Expert systems have also been used to achieve knowledge integration such
as in the case of coronary-care monitoring. The clinical impact of these expert
systems has not been remarkable due to the diculties associated with the
design and maintenance of a complete knowledge base. Model-based systems,
however, represent an alternative to these problems because they allow e-
cient integration of deeper knowledge regarding the underlying physiologi-
cal phenomena being monitored. These systems are specifically designed for
cardiac rhythm interpretation. The cardiac arrhythmia recognition by model-
based ECG matching (CARMEM) has the ability to provide online param-
eter adaptation to simulate complex rhythms and to match observed ECG
signals (Hernndez et al., 2000). It is hoped that this model could be useful for
explanation of the origin of cardiac arrhythmias and provide a contribution
toward their robust characterization of coronary-care units.
4.8 Concluding Remarks
This chapter examined applications of CI techniques for the diagnoses of
CVDs by processing the ECG signal. This field has matured to the extent
where most of the algorithms available have been tried and tested on ECG
recordings. The trend now is toward producing smaller and more portable
ECG devices incorporating PDA technology and other technologies such as
wireless communication. Motivation in commercialization of ECG devices
stems from several evolving factors, which include improvements in the health
care system and response times, increase in awareness of one's personal well-
being, and the need for personal health-monitoring devices. From this per-
spective, there is an emerging requirement for automated heart-monitoring
systems, which require better intelligence to be incorporated as opposed to
the era where heart monitoring was reserved for cardiologists. Nevertheless,
the fundamental challenges in the design of automated systems remain, the
most prominent being the need to develop more accurate CI techniques to aid
Search WWH ::
Custom Search